Re: Abstract Grob property storage into Mutable_properties. (issue 561640043 by address@hidden)

[hanwenn]

On 2020/04/13 20:10:35, dak wrote:
>
https://codereview.appspot.com/561640043/diff/565900043/lily/include/mutable-properties.hh
> File lily/include/mutable-properties.hh (right):
> 
>
https://codereview.appspot.com/561640043/diff/565900043/lily/include/mutable-properties.hh#newcode31
> lily/include/mutable-properties.hh:31: class Iterator {
> On 2020/04/13 17:22:52, hanwenn wrote:
> > On 2020/04/13 17:15:19, Dan Eble wrote:
> > > This is useful, but instead of choosing your own method names, why
not give
> it
> > > an interface consistent with standard iterators (where the
semantics match)?
> > > 
> > > https://en.cppreference.com/w/cpp/named_req/InputIterator
> > > 
> > > I'd leave out the key() and value() methods and let it dealing
with just a
> > > generic list of SCM values.
> > > 
> > > If you implemented enough of the iterator interface, and a class
to wrap the
> > raw
> > > SCM list into an object addressable with begin() and end(), I
think you
> would
> > be
> > > able to write range-for statements something like this:
> > > 
> > > for (SCM assoc : ly_list(my_alist))
> > >   {
> > >     SCM key = scm_car(assoc);
> > >     SCM val = scm_cdr(assoc);
> > >     ...;
> > >   }
> > > 
> > > If you're not interested in doing this, I might try it myself.
> > 
> > By structuring it like this, you enforce the implementation to store
the
> > key/value as SCM cells, which is exactly what we want to get away
from.
> 
> I don't see that there is a plan to get away from the value being an
SCM cell,
> just the key (at least explicitly).  Am I overlooking something?

The value itself is not necessarily a cell. You can have #t as a value
(I am sure you know that.)

Making the Key/value combination a cell forces one to store them as
such, or to construct a cell on the fly. 
There is an experiment where I store keys as uint16. 

https://codereview.appspot.com/561640043/